1Costello H, et al. BMJ Open 2019;9:e027925. doi:10.1136/bmjopen-2018-027925

Open access

Systematic review and meta-analysis of the association between peripheral inflammatory cytokines and generalised anxiety disorder

Harry Costello, Rebecca L Gould, Esha Abrol, Robert Howard

To cite: Costello H, Gould RL, Abrol E, et al. Systematic review and meta-analysis of the association between peripheral inflammatory cytokines and generalised anxiety disorder. BMJ Open 2019;9:e027925. doi:10.1136/bmjopen-2018-027925

► Prepublication history and additional material for this paper are available online. To view these files, please visit the journal online (http:// dx. doi. org/ 10. 1136/ bmjopen- 2018- 027925).

Received 14 November 2018Revised 3 April 2019Accepted 30 May 2019

Division of Psychiatry, University College London, London, UK

Correspondence toDr Harry Costello; harry. costello@ ucl. ac. uk

Research

© Author(s) (or their employer(s)) 2019. Re-use permitted under CC BY. Published by BMJ.

AbstrACtObjective Inflammation has been implicated in the aetiology of mental illness. We conducted the first systematic review and meta-analysis of the association between peripheral markers of inflammation and generalised anxiety disorder (GAD).Design Systematic review and meta-analysis of studies measuring peripheral cytokine levels in people with GAD compared with controls.Data sources MEDLINE (1950–), EMBASE (1947–), PsycINFO (1872–) and Web of Science (1945–) databases up until January 2018.Eligibility criteria Primary, quantitative research studies of people with a diagnosis of GAD assessed using a standardised clinical interview that measured peripheral inflammatory markers.Data extraction and synthesis Two independent reviewers extracted data and assessed study quality. Meta-analysis using a random-effects model was conducted for individual cytokines where data from three or more studies were available.results 14 of 1718 identified studies met the inclusion criteria, comprising 1188 patients with GAD and 10 623 controls. In total 16 cytokines were evaluated. Significantly raised levels of C reactive protein (CRP), interferon-γ and tumour necrosis factor-α were reported in patients with GAD compared with controls in two or more studies. Ten further proinflammatory cytokines were reported to be significantly raised in GAD in at least one study. However, 5 of 14 studies found no difference in the levels of at least one cytokine. Only CRP studies reported sufficient data for meta-analysis. CRP was significantly higher in people with GAD compared with controls, with a small effect size (Cohen’s d=0.38, 0.06–0.69), comparable with that reported in schizophrenia. However, heterogeneity was high (I2=75%), in keeping with meta-analyses of inflammation in other psychiatric conditions and reflecting differences in participant medication use, comorbid depression and cytokine sampling methodology.Conclusion There is preliminary evidence to suggest an inflammatory response in GAD, but it remains unclear whether inflammatory cytokines play a role in the aetiology. GAD remains a poorly studied area of neuroinflammation compared with other mental disorders, and further longitudinal studies are required.

IntrODuCtIOnThere is growing evidence for immune-me-diated pathogenic mechanisms in several psychiatric disorders with discrete profiles of inflammatory mechanisms.1 Epidemiolog-ical evidence has shown an increased risk of mood disorders and psychosis in people with a history of severe infection or autoimmune conditions.2 3 This has been supported by genome-wide association studies implicating multiple immune signalling pathways,4 and altered profiles of proinflammatory cytokines and acute phase reactants in schizophrenia,5 depression,6 obsessive compulsive disorder (OCD)7 and bipolar disorder.8 However, the relationship between inflammation and mental illness remains poorly understood and controversial, with a number of proposed potential neuropathological mechanisms,9 10 including changes in microglial function,1 glutamatergic excitotoxicity,11 synaptic plasticity12 and reduced hippocampal neurogenesis.13

Despite increasing interest in the role of inflammation in mental illness, relatively little research has focused on potential associations

strengths and limitations of this study

► This is the first study to conduct a comprehensive systematic review and meta-analysis of periph-eral inflammatory markers in generalised anxiety disorder.

► A wide range of databases were searched and a large number of papers screened for inclusion in the study, 14 of which were subjected to quality assess-ment and detailed critical appraisal.

► It was only possible to conduct a meta-analysis of C reactive protein, and it was not possible to exam-ine publication bias due to the limited number of studies identified for inclusion in the meta-analysis.

► The high levels of heterogeneity across studies mean that findings should be interpreted with caution.

on Novem

ber 6, 2020 by guest. Protected by copyright.

http://bmjopen.bm

j.com/

BM

J Open: first published as 10.1136/bm

jopen-2018-027925 on 19 July 2019. Dow

nloaded from

2 Costello H, et al. BMJ Open 2019;9:e027925. doi:10.1136/bmjopen-2018-027925

Open access

with anxiety disorders.14 These are common, with an esti-mated lifetime prevalence of 7.3%–28.8%, are associated with substantial functional impairment and are estimated to cost between $42 and $47 billion to the US economy each year.15 16 However, only 60% of patients are thought to respond to pharmacological and psychological treat-ments, and understanding of the underlying pathophysi-ological mechanisms of anxiety disorders remains poor.17

Generalised anxiety disorder (GAD) is the most common anxiety disorder, with a degree of associated disability equivalent to that of major depressive disorder (MDD).18 Despite psychopharmacological19 and psycho-logical20 treatments showing effectiveness in GAD, 42% of people living with GAD experience ongoing symp-toms after 12 years, and half of remitted patients experi-ence recurrence.18 GAD is more prevalent in those with inflammatory conditions such as rheumatoid arthritis (RA),21 22 with case series studies suggesting symptoms are less common with immune-modulating treatment targeting specific inflammatory cytokines.23 The chronic clinical course and relatively high probability of recur-rence in GAD, in addition to preliminary evidence of an inflammatory component in other anxiety disorders,7 24 suggest that inflammation could be an important neuro-biological mechanism in the aetiology of this disorder.

To date, two previous reviews of inflammatory biomarkers in GAD have been conducted. Of these, however, one was a narrative review25 and the other was restricted to literature published within the last decade,14 and with a focus on all anxiety disorders. Both reviews reported that there was preliminary evidence for inflam-matory changes in GAD. However, only three studies were identified by the systematic review reporting cytokine changes in GAD and no meta-analysis was performed. No study to date has conducted a comprehensive systematic review and meta-analysis of all current literature focusing on GAD or commented on the longitudinal association between inflammation and GAD.

We aimed to systematically review the cross-sectional and longitudinal associations between inflammatory biomarkers and GAD, and perform the first meta-analysis of inflammatory biomarkers in GAD.

MEthODsWe conducted a systematic review of studies that had included people with GAD who had undergone peripheral cytokines measurement and a between-group meta-anal-ysis of cytokine levels in people with GAD compared with controls. We conducted the study according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines.26

We searched MEDLINE (1950–), EMBASE (1947–), PsycINFO (1872–) and Web of Science (1945–) databases up until January 2018. Reference lists of eligible studies were then searched for further ones that met the eligi-bility criteria.

Our search terms (see online supplementary appendix 1 for further details) were the following: (inflammat* or cytokine or interferon or IFN or interleukin or ‘trans-locator protein’ or TSPO or ‘tumour necrosis factor’ or ‘tumor necrosis factor’ or TNF or IL-1 or IL-2 or IL-4 or IL-7 or IL-6 or IL-8 or IL-10 or migroglia or t-cell or lymphocyte or ‘C-reactive protein’ or ‘C reactive protein’ or CRP or ‘acute phase protein’ or ‘fibrinogen’) and (‘generalised anxiety disorder’ or ‘generalized anxiety disorder’ or GAD or worry).

We included primary, quantitative research studies (including unpublished theses and dissertations), written in any language, that included people with a diagnosis of GAD assessed using standardised clinical interview (eg, Structured Clinical Interview for DSM27) or standardised psychometric instruments. Studies reported cross-sec-tional or longitudinal data in clinical or community populations. Cross-sectional studies measured inflamma-tory biomarker concentrations in anxious people versus non-anxious healthy controls, while longitudinal studies measured inflammatory biomarker concentrations at baseline and anxiety scores at follow-up. Inflammatory markers were measured in the unstimulated state (no antigen-induced stimulation of cytokine production) and sampled from peripheral blood, cerebrospinal fluid (CSF) or saliva at any time of day. Exclusion criteria included studies with less than five participants, studies on animals and studies where subjects were participants in the treatment arms of clinical trials.

Patient and public involvementThere was no patient or public involvement in the study.

Data extraction and quality assessmentData were extracted and quality assessed for all studies that met the eligibility criteria by two independent raters (HC, EA), with disagreements settled by consensus and discussion. For each cytokine, we extracted the means, variance estimates or 95% CIs and sample size for GAD and control groups. We also extracted demographic data (eg, age, sex) and clinical data (eg, medication use, comorbid depression, severity) where available. Authors were contacted for further information, where necessary.

Risk of bias and study quality were evaluated using the Newcastle-Ottawa Quality Assessment Scale.28 Other potential confounding factors (including assay type and sensitivity, inflammatory marker analysis and recruitment methods) were also examined to allow more detailed bias and quality analysis of studies.

strategy for data synthesisSeparate meta-analyses were performed for individual biomarkers in GAD versus controls if sufficient data were available from a minimum of three studies. Due to different measurement methods and anticipated high heterogeneity, we estimated a standardised mean differ-ence (SMD) for each inflammatory marker and used a random-effects model for meta-analysis, conducted using

on Novem

ber 6, 2020 by guest. Protected by copyright.

http://bmjopen.bm

j.com/

BM

J Open: first published as 10.1136/bm

jopen-2018-027925 on 19 July 2019. Dow

nloaded from

3Costello H, et al. BMJ Open 2019;9:e027925. doi:10.1136/bmjopen-2018-027925

Open access

RevMan V.5. Heterogeneity across studies was quantified with the I2 statistic, with a value of 25% typically regarded as low, 50% as medium and 75% as high.29 If studies were longitudinal or trials of interventions with multiple data collection points, we examined baseline data only to avoid skewed meta-analysis from inclusion of more than one effect size from the same study.

rEsultssystematic reviewWe identified 1718 papers, excluded 1598 of these by titles and abstracts, and retrieved the remaining 120 papers, of which 14 met the eligibility criteria and were included in the final systematic review (see figure 1).

The primary reasons for rejection were that no diagnosis of GAD was recorded or no inflammatory marker was measured.

The characteristics of the 14 included studies are shown in tables 1 and 2. Studies comprised a total of 1188 people with a diagnosis of GAD and 10 623 controls, with a further 116 participants from a study that did not report GAD and control group sizes.

In total, 16 different cytokines were evaluated (see table 2). C reactive protein (CRP) (9/14 studies, 64.2%), tumour necrosis factor-α (TNF-α) (6/14 studies, 42.9%), interleukin-6 (IL-6) (5/14 studies, 35.7%) and inter-feron-γ (IFN-γ) (3/14 studies, 21.4%) were the most commonly studied. All other cytokines were only analysed in two or less studies.

Figure 1 Flow of studies in the systematic review and meta-analysis. GAD, generalised anxiety disorder. on Novem

ber 6, 2020 by guest. Protected by copyright.

http://bmjopen.bm

j.com/

BM

J Open: first published as 10.1136/bm

jopen-2018-027925 on 19 July 2019. Dow

nloaded from

4 Costello H, et al. BMJ Open 2019;9:e027925. doi:10.1136/bmjopen-2018-027925

Open access

Tab

le 1

S

tud

y an

d c

linic

al c

hara

cter

istic

s

Stu

dy

Co

untr

yS

tud

y ty

pe

Infl

amm

ato

ry

mar

kers

NS

tand

ard

ised

d

iag

nost

ic

asse

ssm

ent/

crit

era

A

nxie

ty

mea

sure

Ag

e (y

ears

, SD

)%

Fem

ale

BM

I

Cur

rent

sm

oki

ngM

edic

ated

Phy

sica

l hea

lth

com

orb

idit

ies

Men

tal h

ealt

h co

mo

rbid

itie

sG

AD

Co

ntro

lG

AD

Co

ntro

lG

AD

Co

ntro

lG

AD

Co

ntro

l

Ban

kier

et

al38

US

AC

ase–

cont

rol

CR

P15

30S

CID

, DS

M-I

VN

RN

R67

.6

(12.

7)N

R33

%N

RN

RG

AD

: NR

Con

trol

s: 1

0%N

RA

ll p

artic

ipan

ts

had

CV

D.

Exc

lud

ed o

ther

co

nditi

ons.

Exc

lud

ed.

Cop

elan

d

et a

l39U

SA

Coh

ort,

p

rosp

ectiv

eC

RP

146

5664

Chi

ld a

nd

Ad

oles

cent

P

sych

iatr

ic

Ass

essm

ent 

<16

, Yo

ung

Ad

ult

Psy

chia

tric

A

sses

smen

t >

16,

DS

M-I

V c

riter

ia

Tota

l num

ber

of

anx

iety

sy

mp

tom

s (r

ange

: 0–6

)

14.2

1, a

ll su

bje

cts

(od

ds

or m

eans

ra

tio 1

.12

(1.0

3–1.

22) w

ith G

AD

dx)

48.7

% (o

dd

s or

m

eans

rat

io 2

.02

(1.0

4–3.

92) w

ith

GA

D d

x)

22.3

7 (5

.62)

(od

ds

or m

eans

rat

io

1.08

(1.0

5–1.

11)

with

GA

D d

x)

Tota

l sam

ple

: 13

.5%

(od

ds/

mea

ns r

atio

w

ith G

AD

: 2.8

6)

30.2

% ‘u

se

med

icat

ion’

(o

dd

s/m

eans

ra

tio w

ith

GA

D: 2

.00)

34.7

% h

ad

‘rec

ent

heal

th

ailm

ents

’.

Tota

l sa

mp

le: 3

9.9%

co

mor

bid

MD

D.

De

Ber

ard

is

et a

l42C

anad

aC

ross

-se

ctio

nal

CR

P70

No

cont

rol

SC

ID, D

SM

-IV

HA

M-A

(s

core

 >20

for

incl

usio

n)

28.2

(5

.3)

–51

.40%

–22

.1

(1.6

7)–

NR

Exc

lud

ed.

Exc

lud

ed.

Tota

l sam

ple

: 44

.3%

of

par

ticip

ants

had

al

exith

ymia

.E

xclu

ded

oth

er

com

orb

id m

enta

l ill

ness

.

Hog

e et

al44

US

AR

CT

TNF-

α, IL

-670

No

cont

rol

SC

ID, D

SM

-IV

NR

39.1

2–

45.7

0%–

NR

–N

RE

xclu

ded

.E

xclu

ded

.G

AD

: 14.

3%

com

orb

id M

DD

. E

xclu

ded

oth

er

com

orb

id m

enta

l ill

ness

.

Hou

et

al47

UK

Cas

e–co

ntro

lIL

-4, I

L-10

, TN

F-α,

IFN

-γ54

64M

INI,

DS

M-

IV, I

CD

-10

crite

riaH

AD

S, G

AD

-7

(sco

re >

10 fo

r in

clus

ion)

35.0

6 (1

4.45

)25

.75

(8.8

7)34

%50

%24

.84

(5.7

0)22

.45

(3.2

7)G

AD

: 22%

Con

trol

s: 3

4%G

AD

: 67

% u

se

‘anx

ioly

tic’

med

icat

ion.

E

xclu

ded

ot

her

med

icat

ion

use.

Exc

lud

ed.

Exc

lud

ed.

Kha

ndak

er

et a

l33U

KC

ohor

t,

pro

spec

tive

CR

P26

3392

DA

WB

A, D

SM

-IV

cr

iteria

DA

WB

A15

.56

(0.2

4)15

.53

(0.3

1)90

%52

.30%

22.5

5 (3

.52)

21.4

0 (3

.63)

NR

NR

NR

GA

D: 3

0.77

%

com

orb

id M

DD

. E

xclu

ded

oth

er

com

orb

id m

enta

l ill

ness

.

Kor

keila

et

al40

Finl

and

Cro

ss-

sect

iona

lC

RP,

TN

F-α,

IF

N-γ

116

MIN

IN

RN

RN

R10

0%10

0%A

ll 25

.3

(5.0

)N

RE

xclu

ded

.N

RN

RN

R

Nay

ek a

nd

Gho

sh34

Ind

iaC

ase–

cont

rol

CR

P50

50IC

D-1

0N

R37

.96

(10.

7)37

.00

(12.

08)

54%

23%

NR

NR

Exc

lud

ed.

Exc

lud

ed if

us

ing

HR

T or

O

CP.

Oth

er

med

icat

ions

no

t re

por

ted

.

Exc

lud

ed.

NR

Ogł

odek

et

al48

Pol

and

Cas

e–co

ntro

lS

DF-

1, C

CL-

5,

MC

P-1

120

40D

SM

-VN

R41

.4

(3.5

)40

.8

(3.1

)50

%N

RN

RN

RN

RE

xclu

ded

.E

xclu

ded

.A

ll p

artic

ipan

ts

had

com

orb

id

per

sona

lity

dis

ord

er.

Exc

lud

ed o

ther

co

mor

bid

men

tal

illne

ss.

Con

tinue

d

on Novem

ber 6, 2020 by guest. Protected by copyright.

http://bmjopen.bm

j.com/

BM

J Open: first published as 10.1136/bm

jopen-2018-027925 on 19 July 2019. Dow

nloaded from

5Costello H, et al. BMJ Open 2019;9:e027925. doi:10.1136/bmjopen-2018-027925

Open access

Stu

dy

Co

untr

yS

tud

y ty

pe

Infl

amm

ato

ry

mar

kers

NS

tand

ard

ised

d

iag

nost

ic

asse

ssm

ent/

crit

era

A

nxie

ty

mea

sure

Ag

e (y

ears

, SD

)%

Fem

ale

BM

I

Cur

rent

sm

oki

ngM

edic

ated

Phy

sica

l hea

lth

com

orb

idit

ies

Men

tal h

ealt

h co

mo

rbid

itie

sG

AD

Co

ntro

lG

AD

Co

ntro

lG

AD

Co

ntro

lG

AD

Co

ntro

l

Tang

et

al35

Chi

naC

ase–

cont

rol

CR

P, IL

-1α,

IL

-2, I

L-5,

IL-6

, IL

-8, I

L-12

p70

, IF

N-γ

, GM

-C

SF

4848

MIN

I, D

SM

-IV

GA

D-7

, SA

I, TA

I40

.75

(12.

21)

39.5

6 (1

0.06

)58

.33%

64.1

7%22

.56

(2.7

3)22

.69

(2.6

3)G

AD

: 29%

Con

trol

s: 2

3%E

xclu

ded

.E

xclu

ded

acu

te

illne

ss.

Chr

onic

co

mor

bid

ities

N

R.

Exc

lud

ed.

Tofa

ni e

t al

46Ita

lyC

ase–

cont

rol

IL-2

, IL-

1014

10M

INI,

DS

M-I

VG

AD

-7N

RN

RN

RN

RN

RN

RN

RE

xclu

ded

.N

RE

xclu

ded

.

Voge

lzan

gs

et a

l36H

olla

ndC

ohor

tC

RP,

IL-6

, TN

F-α

454

556

CID

I, D

SM

crit

eria

BA

ITo

tal s

amp

le: 4

1.8

(13.

1)66

.90%

25.6

(5.1

)To

tal s

amp

le:

38.2

%N

RTo

tal s

amp

le:

6.2%

CV

D,

4.9%

dia

bet

es,

mea

n of

0.4

ot

her

chro

nic

dis

ease

s.

Tota

l sam

ple

: 58

.4%

com

orb

id

MD

D.

Exc

lud

ed o

ther

co

mor

bid

men

tal

illne

ss.

Yang

et

al45

Chi

naC

ase–

cont

rol

IL-1

, IL-

4,

TNF-

α28

41M

INI,

DS

M-I

VH

AM

-A55

.1

(6.9

)55

.9

(5.6

)53

.60%

48.8

0%22

.0

(4.4

)22

.5

(3.7

)G

AD

: 35.

7%C

ontr

ol: 2

4.4%

Exc

lud

ed.

Ad

diti

onal

gro

up

with

com

orb

id

asth

ma.

Exc

lud

ed o

ther

co

mor

bid

ities

.

Exc

lud

ed.

Zah

m41

US

AC

ohor

t,

pro

spec

tive

CR

P, IL

-6,

TNF-

α93

728

CD

IS, D

SM

-IV

NR

68 (9

.6)

17%

NR

NR

NR

NR

All

pat

ient

s ha

d

hist

ory

of C

VD

.G

AD

: 60.

0%

com

orb

id M

DD

, 86

.2%

had

lif

etim

e hi

stor

y M

DD

.

–, n

ot a

pp

licab

le; B

AI,

Bec

k A

nxie

ty In

vent

ory;

CC

L-5,

 che

mok

ine

C-C

mot

if lig

and

5; C

DIS

, Com

put

eriz

ed D

iagn

ostic

Inte

rvie

w S

ched

ule;

CID

I, C

omp

osite

Inte

rvie

w D

iagn

ostic

Inst

rum

ent;

CR

P, C

rea

ctiv

e p

rote

in; C

VD

, car

dio

vasc

ular

dis

ease

; D

AW

BA

, Dev

elop

men

t an

d W

ell-

Bei

ng A

sses

smen

t; D

SM

, Dia

gnos

tic a

nd S

tatis

tical

Man

ual o

f Men

tal D

isor

der

s; d

x, d

iagn

osis

; GA

D, g

ener

alis

ed a

nxie

ty d

isor

der

; GA

D-7

, Gen

eral

ised

Anx

iety

Dis

ord

er A

sses

smen

t; G

M-C

SF,

gra

nulo

cyte

-m

acro

pha

ge c

olon

y-st

imul

atin

g fa

ctor

; HA

DS

, Hos

pita

l Anx

iety

and

Dep

ress

ion

Sca

le; H

AM

-A, H

amilt

on A

nxie

ty R

atin

g S

cale

; HR

T, h

orm

one

rep

lace

men

t th

erap

y; IC

D, I

nter

natio

nal C

lass

ifica

tion

of D

isea

ses;

IFN

-γ, i

nter

fero

n-γ;

IL, i

nter

leuk

in;

MC

P-1

, mon

ocyt

e ch

emoa

ttra

ctan

t p

rote

in-1

; MD

D, m

ajor

dep

ress

ive

dis

ord

er; M

INI,

Min

i-In

tern

atio

nal N

euro

psy

chia

tric

Inte

rvie

w; N

R, n

ot r

epor

ted

; OC

P, o

ral c

ontr

acep

tive 

pill

; RC

T, r

and

omis

ed c

ontr

olle

d t

rial;

SA

I, S

tate

Anx

iety

Inve

ntor

y;

SC

ID, S

truc

ture

d C

linic

al In

terv

iew

for

DS

M-I

V; S

DF-

1, s

trom

al d

eriv

ed fa

ctor

-1; T

AI,

Test

Anx

iety

Inve

ntor

y; T

NF-

α, t

umou

r ne

cros

is fa

ctor

-α.

Tab

le 1

C

ontin

ued

on Novem

ber 6, 2020 by guest. Protected by copyright.

http://bmjopen.bm

j.com/

BM

J Open: first published as 10.1136/bm

jopen-2018-027925 on 19 July 2019. Dow

nloaded from

6 Costello H, et al. BMJ Open 2019;9:e027925. doi:10.1136/bmjopen-2018-027925

Open access

Tab

le 2

In

flam

mat

ory

mar

ker

sam

plin

g an

d a

naly

sis

Stu

dy

GA

DC

ont

rol

Infl

amm

ato

ry

mar

kers

Nat

ure

of

sam

ple

Cro

ss-s

ecti

ona

l o

r lo

ngit

udin

alT

ime

of

day

of

sam

ple

Fast

ed

per

iod

b

efo

re

sam

ple

Ass

ay m

etho

d

Ass

ay

sens

itiv

ity

rep

ort

ed

Infl

amm

ato

ry

mar

ker

cut-

off

us

edC

onf

oun

din

g f

acto

rs

cont

rolle

d f

or

Ban

kier

et

al38

1530

CR

PB

lood

Cro

ss-s

ectio

nal

NR

NR

Hig

h se

nsiti

vity

tu

rbid

imet

ric

imm

unoa

ssay

Yes

CR

P >

3 m

g/L

for

sign

ifica

nce

Age

, sex

, ed

ucat

ion,

M

DD

, ob

esity

, sm

okin

g hi

stor

y,

typ

e II

dia

bet

es

mel

litus

, hyp

erte

nsio

n,

hyp

erlip

idae

mia

, oth

er

men

tal i

llnes

s.

Cop

elan

d e

t al

3914

656

64C

RP

Who

le

blo

od

spot

s

Long

itud

inal

: sa

mp

led

age

d

9–16

, 19

and

21

yea

rs o

ld

NR

NR

Bio

tin-

stre

pta

vid

in-b

ased

im

mun

ofluo

rom

etric

sy

stem

Yes

Exc

lud

ed

if >

10 m

g/L

Age

, sex

, rac

e, S

ES

, B

MI,

med

icat

ion

use,

sub

stan

ce u

se,

rece

nt p

hysi

cal i

llnes

s,

chro

nic

illne

ss.

De

Ber

ard

is

et a

l4270

No

cont

rol

CR

PS

erum

Cro

ss-s

ectio

nal

07:0

0–08

:30

10-h

our

fast

Hig

hly

sens

itive

ne

phe

lom

etric

ass

ayYe

sN

oA

ge, s

ex, B

MI,

MD

D,

phy

sica

l illn

ess,

ot

her

men

tal i

llnes

s,

med

icat

ion

use

Hog

e et

al44

70N

o co

ntro

lTN

F-α,

IL-6

Pla

sma

Long

itud

inal

: sa

mp

led

p

rep

sych

olog

ical

an

d

pos

tpsy

chol

ogic

al

inte

rven

tion

13:0

0–16

:30

NR

NR

No

No

Age

, sex

, eth

nici

ty,

MD

D, m

edic

atio

n us

e,

phy

sica

l illn

ess,

oth

er

men

tal i

llnes

s

Hou

et

al47

5464

IL-4

, IL-

10,

TNF-

α, IF

N-γ

Ser

umC

ross

-sec

tiona

l09

:00–

10:0

0N

RM

ultip

lex

ultr

asen

sitiv

e im

mun

oass

ay

Yes

No

Age

, sex

, BM

I, sm

okin

g, a

lcoh

ol

cons

ump

tion,

MD

D,

phy

sica

l illn

ess,

oth

er

men

tal i

llnes

s.

Kha

ndak

er

et a

l3326

3392

CR

PS

erum

Cro

ss-s

ectio

nal

NR

‘Ove

rnig

ht’

Aut

omat

ed

par

ticle

-enh

ance

d

imm

unot

urb

idim

etric

as

say

No

Exc

lud

ed

if >

10 m

g/L

Age

, sex

, par

enta

l S

ES

, eth

nici

ty,

mat

erna

l age

at

del

iver

y,

conc

urre

nt in

fect

ion,

fa

mily

his

tory

of

infla

mm

ator

y d

isea

se,

MD

D.

Kor

keila

et

al40

116

CR

P, T

NF-

α,

IFN

-γB

lood

Cro

ss-s

ectio

nal

NR

NR

NR

No

No

BM

I.

Nay

ek a

nd

Gho

sh34

5050

CR

PS

erum

Cro

ss-s

ectio

nal

NR

NR

Par

ticle

-enh

ance

d

turb

idim

etric

im

mun

oass

ay

tech

niq

ue

No

Exc

lud

ed if

‘r

aise

d E

SR

’A

ge, s

ex, S

ES

, re

ligio

n, m

arita

l sta

tus,

lo

calit

y, B

MI >

30,

phy

sica

l illn

ess.

Con

tinue

d

on Novem

ber 6, 2020 by guest. Protected by copyright.

http://bmjopen.bm

j.com/

BM

J Open: first published as 10.1136/bm

jopen-2018-027925 on 19 July 2019. Dow

nloaded from

7Costello H, et al. BMJ Open 2019;9:e027925. doi:10.1136/bmjopen-2018-027925

Open access

Stu

dy

GA

DC

ont

rol

Infl

amm

ato

ry

mar

kers

Nat

ure

of

sam

ple

Cro

ss-s

ecti

ona

l o

r lo

ngit

udin

alT

ime

of

day

of

sam

ple

Fast

ed

per

iod

b

efo

re

sam

ple

Ass

ay m

etho

d

Ass

ay

sens

itiv

ity

rep

ort

ed

Infl

amm

ato

ry

mar

ker

cut-

off

us

edC

onf

oun

din

g f

acto

rs

cont

rolle

d f

or

Ogł

odek

et

al48

120

40S

DF-

1, C

CL-

5, M

CP

-1P

lasm

aC

ross

-sec

tiona

l07

:00–

09:0

0Fa

sted

, d

urat

ion

NR

ELI

SA

Yes

No

Sex

, oth

er m

enta

l ill

ness

, phy

sica

l ill

ness

, sub

stan

ce

mis

use,

sm

okin

g st

atus

, med

icat

ion

use.

Tang

et

al35

4848

CR

P, IL

-1α,

IL

-2, I

L-5,

IL

-6, I

L-8,

IL-

12p

70, I

FN-γ

, G

M-C

SF

Ser

umC

ross

-sec

tiona

l09

:00–

10:0

0N

RE

LIS

AN

oN

oA

ge, s

ex, e

duc

atio

n,

BM

I, sm

okin

g st

atus

, al

coho

l con

sum

ptio

n,

acut

e p

hysi

cal i

llnes

s,

othe

r m

enta

l illn

ess,

m

edic

atio

n us

e.

Tofa

ni e

t al

4614

10IL

-2, I

L-10

Pla

sma

Cro

ss-s

ectio

nal

NR

NR

Imm

unoe

nzym

atic

as

say

No

No

Med

icat

ion

use.

Voge

lzan

gs

et a

l3645

455

6C

RP,

IL-6

, TN

F-α

Pla

sma

Cro

ss-s

ectio

nal

08:0

0–09

:00

‘Ove

rnig

ht’

ELI

SA

Yes

No

Age

, sex

, ed

ucat

ion,

sm

okin

g st

atus

, al

coho

l int

ake,

p

hysi

cal a

ctiv

ity,

BM

I, p

hysi

cal i

llnes

s,

med

icat

ion

use,

MD

D,

othe

r m

enta

l illn

ess.

Yang

et

al45

2841

IL-1

, IL-

4, IL

-6,

TN

F-α

Sal

iva

Cro

ss-s

ectio

nal

NR

‘Ove

rnig

ht’

ELI

SA

Yes

No

Age

, sex

, sm

okin

g st

atus

, BM

I, m

edic

atio

n us

e,

phy

sica

l illn

ess,

oth

er

men

tal i

llnes

s.

Zah

m41

9372

8C

RP,

IL-6

, TN

F-α

Ser

umC

ross

-sec

tiona

lN

o (fa

stin

g,

dur

atio

n N

R)

Fast

ed,

dur

atio

n N

RE

LIS

AYe

sN

oA

ge, s

ex, S

ES

, BM

I, ill

icit

sub

stan

ce u

se,

alco

hol u

se, s

mok

ing

stat

us, p

hysi

cal

activ

ity, p

hysi

cal

illne

ss.

BM

I, b

ody

mas

s in

dex

; CC

L-5,

che

mok

ine

C-C

mot

if lig

and

5; C

RP,

C r

eact

ive

pro

tein

; ES

R, e

ryth

rocy

te s

edim

enta

tion

rate

; GA

D, g

ener

alis

ed a

nxie

ty d

isor

der

; GM

-CS

F, g

ranu

locy

te-

mac

rop

hage

col

ony-

stim

ulat

ing

fact

or; I

FN-γ

, int

erfe

ron-

γ; IL

, int

erle

ukin

; MC

P-1

, mon

ocyt

e ch

emoa

ttra

ctan

t p

rote

in-1

; MD

D, m

ajor

dep

ress

ive

dis

ord

er; N

R, n

ot r

epor

ted

; SD

F-1,

str

omal

d

eriv

ed fa

ctor

-1; S

ES

, soc

ioec

onom

ic s

tatu

s; T

NF-

α, t

umou

r ne

cros

is fa

ctor

-α.

Tab

le 2

C

ontin

ued

on Novem

ber 6, 2020 by guest. Protected by copyright.

http://bmjopen.bm

j.com/

BM

J Open: first published as 10.1136/bm

jopen-2018-027925 on 19 July 2019. Dow

nloaded from

8 Costello H, et al. BMJ Open 2019;9:e027925. doi:10.1136/bmjopen-2018-027925

Open access

Twelve studies (85.7%) reported the assay method used, all of which were versions of an ELISA or enzyme immu-noassay. However, only seven studies (50%) reported assay sensitivity. All but one study used blood component samples to assess inflammatory marker levels, with the most common sample type being serum (n=6, 42.9%) and plasma (n=4, 28.6%).

risk of bias and quality in individual studiesAll included studies had adequate case definition, with participants meeting the diagnostic criteria for GAD according to the Diagnostic and Statistical Manual of Mental Disorders (DSM) or the International Classifica-tion of Diseases, with 12 studies (85.7%) using a struc-tured clinical interview for assessment (see table 1).

Most (71.4%) studies included people aged 18–65, but two studies (14.3%) only included participants over the age of 50 and a further two studies (14.3%) used adoles-cent participant cohorts. The majority (78.6%) of studies accounted for age and sex differences in their analyses. Only 8 of 14 (57.1%) studies recorded participants’ body mass index (BMI), which is known to correlate with inflammation, and only half of these accounted for BMI differences in analysis of group differences30 (see table 2).

Use of psychotropic medication and the presence of comorbid MDD are important moderators of inflam-mation in other psychiatric disorders.24 Six studies (42.8%) excluded patients who used psychiatric or other immune-modulating medication, although only two studies (4.3%) reported medication use. The majority of studies (64.2%) either excluded patients with comorbid MDD or adjusted for this in the analyses.

Concurrent physical illness is clearly an important determinant of inflammatory cytokine levels, and this was accounted for by the majority of included studies by either excluding participants with comorbidities (5 studies, 35.7%) or adjusting for chronic physical illness in group comparisons (6 studies, 42.8%), although two studies specifically only included participants with comorbid cardiovascular disease (CVD). Use of a prede-termined cut-off value for cytokine levels was employed by three studies (21.4%) to ensure that cases with acute infection were excluded from the sample.

Many inflammatory markers exhibit a diurnal pattern of expression and are affected by consumption of food; thus, time of day of sampling and whether the sample was taken in a fasted state are important factors to consider in analysing relative levels of cytokines.31 However, time of day of sampling was only recorded in a minority of studies (6 studies, 42.8%), and the same number of studies recorded whether fasted samples were taken.

The overall quality of studies included in the review varied significantly, with Newcastle-Ottawa Scale scores ranging from 2 to 9 (see table 3). The area in which most studies were inadequate was in reporting non-response rate and detailing recruitment methods (see table 3). Lowest quality studies were abstracts or dissertations, and

two studies lacked control groups as only patients with GAD were sampled (see table 4).

C reactive proteinCRP is a critical early proinflammatory surveillance molecule involved in the activation of the complement system and both innate and adaptive immune systems.32 We identified nine studies that investigated the associa-tion between GAD and CRP, comprising a total of 11 486 participants (see table 5).

Four studies, involving 578 patients with GAD and 4046 controls, provided sufficient information to conduct a meta-analysis of CRP levels in GAD33–36 (see figure 2). This was the only inflammatory marker for which meta-analysis was possible. Meta-analysis showed signifi-cantly raised CRP in GAD compared with controls (SMD 0.38, 95% CI 0.06 to 0.69; Z=2.36, p=0.02). However, there was a large and statistically significant degree of heteroge-neity between studies (Χ2=12.0; df=3; p=0.007; I2=75%). Given the high heterogeneity and inclusion of less than 10 studies in the meta-analysis, we did not have sufficient power to examine publication bias.37 Two out of four studies were of high quality, scoring 9 on the Newcas-tle-Ottawa Scale, and examined large sample sizes33 36 (see table 4). However, in each of the four meta-analysed studies, different assay methods were used and sampling methods varied significantly (see table 2). The lowest quality study to be included in the meta-analysis did not report mental health comorbidities and recruited participants from an inpatient setting.34 There was also a wide range in the age of participants included in the four studies, with the largest study examining CRP levels in adolescents, which would likely contribute to high heterogeneity.

Five studies33–36 38 (n=4669), one of which was conducted in children aged 16 years old33 and two in participants with comorbid heart disease,38 reported significantly higher CRP levels in participants with a diagnosis of GAD. The largest study39 (n=5810) examining CRP in GAD examined CRP levels in children from baseline measure-ment aged 9–16 years to follow-up at age 19–21. This was the only study to examine the longitudinal association between GAD and CRP, and found a bivariate association both cross-sectionally and over time between GAD and elevated CRP; however, this was accounted for by poten-tial covariates, including BMI and medication use. The only study40 to find an inverse correlation between CRP and GAD was conducted in non-smoking women from a study in Finland and did not specify the numbers of participants with a diagnosis of GAD or group differences.

No difference was found in a cohort study (n=821) that used a combined inflammatory index consisting of CRP, IL-6 and TNF-α in 93 patients with a diagnosis of GAD and controls with a history of CVD.41 Subgroup analysis exam-ining differences in individual inflammatory markers was not reported.41

We found two studies35 42 (n=196) that examined the association of severity of GAD symptoms with CRP level.

on Novem

ber 6, 2020 by guest. Protected by copyright.

http://bmjopen.bm

j.com/

BM

J Open: first published as 10.1136/bm

jopen-2018-027925 on 19 July 2019. Dow

nloaded from

9Costello H, et al. BMJ Open 2019;9:e027925. doi:10.1136/bmjopen-2018-027925

Open access

Table 3 Summary inflammatory marker findings in GAD

Study

N

FindingsControls With GAD (n)

CRP

Bankier et al38 30 15 ↑ in GAD with comorbid CVD compared with controls using a dichotomous outcome of CRP cut-off score (CRP >3 mg/L).

Copeland et al39 5664 146 Longitudinal study in adolescents: ↑ bivariate association both cross-sectionally and over time between GAD and elevated CRP, but accounted for by medication use and BMI.

De Berardis et al42 No control 70 ↑ in patients with GAD with comorbid alexithymia and with increased suicidal ideation, no control group.

Khandaker et al33 3392 26 ↑ in children aged 16 years old with GAD compared with controls, remained ↑ after adjusting for covariates.

Korkeila et al40 116 ↓ in non-smoking women with a diagnosis of GAD compared with controls; however, control group was not described.

Nayek and Ghosh34 50 50 ↑ in patients with GAD compared with controls.

Tang et al35 48 48 ↑ in patients with GAD compared with controls and ↑ with increased severity of GAD.

Vogelzangs et al36 556 454 ↑ in patients with GAD compared with controls in unadjusted data obtained from the authors.

Zahm41 728 93 ↔ between those with and without a current GAD diagnosis (p=0.28) or with and without a lifetime GAD diagnosis, using a combined inflammatory index of CRP, IL-6 and TNF-⍺ measurements.

IL-1

Yang et al45 41 28 ↑ sputum IL-1 in patients aged 50–60 years old with GAD compared with controls.

IL-1α

Tang et al35 48 48 ↑ IL-1α in patients with GAD compared with controls and ↑ with increased severity of GAD.

IL-2

Tang et al35 48 48 ↑ in patients with GAD compared with controls (p<0.001) but ↔ with severity of GAD.

Tofani et al46 10 14 ↔ in patients with GAD compared with controls.

IL-4

Hou et al47 64 54 ↔ in patients with GAD compared with controls.

IL-5

Tang et al35 48 48 ↔ in patients with GAD compared with controls, or association with severity of GAD.

IL-6

Hoge et al44 – 70 No control group: RCT of psychological intervention in GAD.

Tang et al35 48 48 ↑ in patients with GAD compared with controls and ↑ with increased severity of GAD.

Vogelzangs et al36 556 454 ↑ in patients with GAD compared with controls in unadjusted data obtained from author, but ↔ between IL-6 and GAD compared with other anxiety disorders.

Yang et al45 41 28 ↑ sputum in patients with GAD aged 50–60 years old compared with controls.

Zahm41 728 93 ↔ between those with and without a current GAD diagnosis or with and without a lifetime GAD diagnosis, using a combined inflammatory index of CRP, IL-6 and TNF-⍺ measurements.

IL-8

Tang et al35 48 48 ↑ in patients with GAD compared with controls and ↑ with increased severity of GAD.

IL-10

Hou et al47 64 54 ↓ in patients with GAD compared with controls, which remained ↓ after adjustment for covariates.

Continued

on Novem

ber 6, 2020 by guest. Protected by copyright.

http://bmjopen.bm

j.com/

BM

J Open: first published as 10.1136/bm

jopen-2018-027925 on 19 July 2019. Dow

nloaded from

10 Costello H, et al. BMJ Open 2019;9:e027925. doi:10.1136/bmjopen-2018-027925

Open access

One found a significant positive correlation between CRP level and Generalised Anxiety Disorder Assess-ment (GAD-7) scores,35 and the other reporting CRP differences in 70 patients with GAD with and without a diagnosis of alexithymia found a significant association between higher CRP and suicidal ideation.42

Although the meta-analysis and the majority of included studies reported raised CRP in GAD, there was wide varia-tion in reporting and adjustment for important potential moderators, including comorbid MDD, use of medica-tions, assay used and time of day of blood collection, all

of which likely contributed to the high degree of hetero-geneity between studies. Of the nine studies to analyse CRP, four (44.4%) did not exclude or adjust for medi-cation use by participants.33 38 40 41 Comorbid MDD was not adjusted for in the analysis by two studies,39 41 one of which was included in the meta-analysis.39 Only three of the nine studies reported time of sample collection35 36 42 or whether this was in a fasted state,36 41 42 and although all studies used a similar assay method, different assay types were used in every study.

Study

N

FindingsControls With GAD (n)

Tofani et al46 10 14 ↑ in GAD compared with controls.

IL-12p70

Tang et al35 48 48 ↑ in patients with GAD compared with controls but ↔ with severity of GAD.

IFN-γ

Hou et al47 64 54 ↑ in patients with GAD compared with controls which remained ↑ after adjustment for covariates.

Korkeila et al40 116 ↓ in non-smoking women with a diagnosis of GAD compared with controls; however, control group was not described.

Tang et al 201735 48 48 ↑ in patients with GAD compared with controls and ↑ with increased severity of GAD.

TNF-α

Hoge et al44 – 70 No control group: RCT of psychological intervention in GAD.

Hou et al 201747 64 54 ↑ in patients with GAD compared with controls which remained ↑ after adjustment for covariates.

Korkeila et al40 116 ↑ in non-smoking women with a diagnosis of GAD compared with controls, although control group was not described.

Vogelzangs et al36 556 454 ↔ in patients with GAD compared with controls, and ↔ between TNF-α and GAD compared with other anxiety disorders.

Yang et al45 41 28 ↑ sputum TNF-α in patients with GAD aged 50–60 years old compared with controls.

Zahm41 728 93 ↔ between those with and without a current GAD diagnosis or with and without a lifetime GAD diagnosis, using a combined inflammatory index of CRP, IL-6 and TNF-⍺ measurements.

CCL-5/RANTES

Ogłodek et al48 40 120 ↑ in men with GAD and comorbid personality disorder compared with controls.

MCP-1

Ogłodek et al48 40 120 ↑ in GAD and comorbid personality disorder compared with controls.

SDF-1

Ogłodek et al48 40 120 ↑ in GAD and comorbid personality disorder compared with controls.

GM-CSF

Tang et al35 48 48 ↑ in patients with GAD compared with controls and ↑ with increased severity of GAD.

↑,statistically significant increase in inflammatory marker in people with GAD compared with controls (p<0.05); ↓,statistically significant decrease in inflammatory marker in people with GAD compared with controls (p<0.05); ↔,no statistically significant difference in inflammatory marker in people with GAD compared with controls (p>0.05).BMI, body mass index; CCL-5, chemokine C-C motif ligand 5; CRP, C reactive protein; CVD, cardiovascular disease; GAD, generalised anxiety disorder; GM-CSF, granulocyte-macrophage colony-stimulating factor; IFN-γ, interferon-γ; IL, interleukin; MCP-1, monocyte chemoattractant protein-1; RANTES, regulated on activation, normal T cell expressed and secreted; RCT, randomised controlled trial; SDF-1, stromal derived factor-1; TNF-α, tumour necrosis factor-α.

Table 3 Continued

on Novem

ber 6, 2020 by guest. Protected by copyright.

http://bmjopen.bm

j.com/

BM

J Open: first published as 10.1136/bm

jopen-2018-027925 on 19 July 2019. Dow

nloaded from

11Costello H, et al. BMJ Open 2019;9:e027925. doi:10.1136/bmjopen-2018-027925

Open access

Tab

le 4

S

tud

y q

ualit

y as

sess

men

t: N

ewca

stle

-Ott

awa

Sca

le

Sel

ecti

on

Co

mp

arab

ility

Exp

osu

re

Ad

equa

te

case

d

efini

tio

nC

ases

re

pre

sent

ativ

eS

elec

tio

n o

f co

ntro

lsD

efini

tio

n o

f co

ntro

ls

Co

mp

arab

ility

o

f d

esig

n an

d

anal

ysis

Asc

erta

inm

ent

of

exp

osu

reS

ame

met

hod

of

asce

rtai

nmen

tN

on-

resp

ons

e ra

teTo

tal s

tars

Ban

kier

et

al38

◊–

◊◊

◊◊◊

◊◊

8

Cop

elan

d e

t al

39◊

◊◊

◊◊◊

◊◊

–8

De

Ber

ard

is e

t al

42◊

–N

AN

AN

A◊

NA

–2

Hog

e et

al44

◊–

NA

NA

NA

◊N

A–

2

Hou

et

al47

◊◊

◊◊

◊◊◊

◊–

8

Kha

ndak

er e

t al

33◊

◊◊

◊◊◊

◊◊

◊9

Kor

keila

et

al40

◊–

◊–

◊–

◊–

4

Nay

ek a

nd G

hosh

34◊

–◊

–◊◊

◊◊

–6

Ogł

odek

et

al48

◊–

◊◊

◊◊–

◊–

6

Tang

et

al35

◊◊

◊◊

◊◊◊

◊–

8

Tofa

ni e

t al

46◊

––

–◊

◊◊

–4

Voge

lzan

gs e

t al

36◊

◊◊

◊◊◊

◊◊

◊9

Yang

et

al45

◊◊

–◊

◊◊◊

◊–

7

Zah

m41

◊–

––

◊◊◊

◊–

5

–, d

id n

ot m

eet

the

crite

ria; ◊

, met

crit

eria

for

allo

catio

n of

poi

nt o

n N

ewca

stle

-Ott

awa

scal

e; ◊

◊ , t

wo

poi

nts

on N

ewca

stle

-Ott

awa

scal

e; N

A, n

ot a

pp

licab

le.

on Novem

ber 6, 2020 by guest. Protected by copyright.

http://bmjopen.bm

j.com/

BM

J Open: first published as 10.1136/bm

jopen-2018-027925 on 19 July 2019. Dow

nloaded from

12 Costello H, et al. BMJ Open 2019;9:e027925. doi:10.1136/bmjopen-2018-027925

Open access

In summary, of the nine studies to have examined differences between GAD and controls, the majority reported raised CRP in GAD and the meta-analysis found significantly raised CRP in GAD with a small effect size. However, there was a wide variation in study methods, including variable adjustment for mediators of inflam-mation such as comorbid MDD, medication use and sampling methods. Only one study examined CRP in GAD longitudinally, reporting a bivariate association accounted for by health-seeking behaviours.39

InterleukinsSeven studies examined the association between interleu-kins and GAD (see table 3). IL-6 is a mediator of T cell and B cell activation and induces acute phase proteins in hepatocytes, among other functions.32 Pharmacolog-ical blockade of IL-6 action is used to treat several auto-immune conditions including RA, and raised IL-6 has been associated with a number of psychiatric conditions including depression, schizophrenia and post-traumatic

Table 5 Additional critical appraisal

Type of publication

Unrepresentative recruitment methods

Unrepresentative demographics

Between-group differences reported

Adjusted for between-group differences

Bankier et al38 Paper Yes, recruited from cardiology clinic

Yes, older cohort due to cardiac comorbidity required

NR NR

Copeland et al39

Paper No Yes, aged 9–21 only Yes Yes

De Berardis et al42

Paper No No (aged 18–45) No control No control

Hoge et al44 Paper Yes, recruited by advert as part of parent RCT

No (aged >18) No control No control

Hou et al47 Paper No No (aged 18–65) Yes Yes

Khandaker et al33

Paper No Yes, aged 16 years old only

Yes Yes

Korkeila et al40 Abstract Yes, recruited from existing study in Finland

Yes, non-smoking women only

No Yes (BMI only)

Nayek and Ghosh34

Paper Yes, inpatients only No (aged 18–65) Yes Yes

Ogłodek et al48 Paper Yes, comorbid personality disorder

No Yes No

Tang et al35 Paper No No (aged 18–60) Yes Yes

Tofani et al46 Abstract Recruitment method not stated

NR NR NR

Vogelzangs et al36

Paper No No (aged 18–65) Yes Yes

Yang et al45 Paper No Yes (aged 50–60) Yes Yes

Zahm41 Dissertation Yes, cardiology patients only

Yes (aged>50) Yes Yes

BMI, body mass index; NR, not reported; RCT, randomised controlled trial.

Figure 2 Random-effects meta-analysis of CRP levels in GAD versus controls. CRP, C reactive protein; GAD, generalised anxiety disorder; Std, standard; IV, inverse variance.

on Novem

ber 6, 2020 by guest. Protected by copyright.

http://bmjopen.bm

j.com/

BM

J Open: first published as 10.1136/bm

jopen-2018-027925 on 19 July 2019. Dow

nloaded from

13Costello H, et al. BMJ Open 2019;9:e027925. doi:10.1136/bmjopen-2018-027925

Open access

stress disorder (PTSD).32 43 We found IL-6 was the most frequently measured interleukin, with five studies (n=2066) examining changes in patients with GAD compared with controls.35 36 41 44 45 The largest study inves-tigated differences between 454 participants with a diag-nosis of GAD and 556 controls from The Netherlands Study of Depression and Anxiety cohort.36 Although analysis was conducted on anxiety disorders as a whole, the mean difference in IL-6 in people with GAD compared with controls obtained through direct communication with the author showed significantly higher levels in GAD. However, it is unclear whether these differences remain significant after adjustment for group differences and no associations were found between IL-6 and participants who had all types of anxiety disorder.36 Two studies35 45 (n=165), one of which used saliva samples,45 reported significantly higher IL-6 in medication-naïve participants with a diagnosis of GAD compared with age-matched and sex-matched healthy controls.

No difference was found in a combined inflamma-tory index consisting of CRP, IL-6 and TNF-α in a study of 93 patients with GAD and comorbid ischaemic heart disease.41 One case-controlled study35 of 48 Chinese outpatients presenting for the first time with a diagnosis of GAD and 48 age-matched, sex-matched and educa-tion-matched controls accounted for all results for IL-1α, IL-5, IL-8 and IL-12p70. This study found significantly higher levels of IL-1α, IL-8 and IL-12p70 in patients with GAD, in addition to higher levels of IL-1α and IL-8 with increased severity of GAD (as measured by the GAD-7 scale), but did not account for chronic physical comorbidi-ties during recruitment or in analysis. Both IL-1α and IL-8 have proinflammatory functions as chemoattractants for leukocytes and haematopoiesis, and have been targeted for treatments in a number of autoimmune conditions.32 However, there was no association between GAD and IL-5, which is thought to predominantly mediate myeloid cell activation and is a target of treatment in asthma.32 The same study35 also examined IL-2, which has a major role in T cell-mediated autoimmune and inflammatory conditions.32 The results showed significantly higher IL-2 in patients with GAD; however, this conflicted with results from a smaller study (n=24) that found no significant difference between medication-naïve patients with GAD and controls, although few details of the characteristics of participants, sampling or analysis were reported in this abstract.46

One study that measured IL-1 using sputum anal-ysis found significantly higher levels in GAD compared with controls in 69 participants recruited from the same Chinese hospital.45 Although IL-1 is proinflammatory, there are differences in function, dependent on the class of IL-1 protein measured which was not reported in this study.45

IL-4 has several proinflammatory functions, including IgE class switching, expression of major histocompati-bility complex (MHC) class II and acts as a survival factor for T and B cells.32 The only study to measure IL-4 found

no differences between 54 patients with GAD recruited from community mental health teams and primary care after controlling for age, sex, BMI, smoking, alcohol consumption and comorbid depression.47 This study47 also investigated IL-10, which was the only cytokine with an anti-inflammatory function to be measured and is involved in immunosuppression of T cell subsets and B cell immunoglobulin production. This found significantly lower levels of IL-10 (OR 0.35, p=0.003) in patients with GAD. However, this opposed findings from a smaller study (n=24) which reported significantly higher levels of IL-10 in patients with GAD compared with controls, although it was not reported whether this association remained significant after controlling for group differences.46

In summary, IL-6 was the most commonly measured interleukin raised in GAD compared with controls in the majority of studies; however, no study examined the longitudinal association with GAD. Other interleukins were examined by relatively few studies that examined small numbers cross-sectionally with mixed findings.

Interferon-γIFN-γ has antiviral roles, including promoting cytotoxic activity, MHC class I and II upregulation, natural killer (NK) cell activation, and is a treatment target in inflam-matory conditions such as Crohn’s disease.32 Three studies investigated IFN-γ levels in GAD (n=330).35 40 47 The largest study (n=118) found higher IFN-γ in patients with GAD from the UK that remained significant after adjustment for age, gender, BMI, smoking, alcohol and comorbid depression, but did not adjust for anxiolytic medication use in analysis.47 This finding was supported by a study of 96 participants which reported higher IFN-γ levels in GAD and a significant positive correla-tion between anxiety severity and IFN-γ.35 Conflicting findings were reported by a Finnish study of 116 partic-ipants, which found significantly lower IFN-γ in patients with GAD. However, the number of participants with a diagnosis of GAD, differences between groups and adjust-ment for potential confounders were not reported.40 In summary, only a few small cross-sectional studies have examined differences in IFN-γ between GAD and control groups, and their findings were mixed.

tumour necrosis factor-αTNF-α has a wide array of roles in host defence, including initiating a strong acute inflammatory response but limiting duration of inflammatory activation, and is the target of blocking monoclonal antibodies in the treat-ment of a wide array of autoimmune conditions including Crohn’s disease and RA.32 Six studies (n=2300) investi-gated TNF-α in GAD, with mixed findings. Three studies (n=303) found TNF-α significantly raised in patients with GAD compared with controls.40 45 47 However, the largest study to measure TNF-α (n=1010) found no difference between participants with GAD and controls, and no correlation between TNF-α and anxiety symptoms.36 This finding was supported by a study of 93 patients

on Novem

ber 6, 2020 by guest. Protected by copyright.

http://bmjopen.bm

j.com/

BM

J Open: first published as 10.1136/bm

jopen-2018-027925 on 19 July 2019. Dow

nloaded from

14 Costello H, et al. BMJ Open 2019;9:e027925. doi:10.1136/bmjopen-2018-027925

Open access

with GAD and comorbid ischaemic heart disease using a combined inflammatory index of CRP, IL-6 and TNF-α which reported no difference compared with controls.41 In summary, although the majority of the studies to measure differences in TNF-α between GAD and controls reported significantly raised levels, these comprised small cross-sectional studies and the largest study reported no difference.

Other cytokinesOne study compared the levels of the proinflammatory cytokines chemokine C-C motif ligand 5 (CCL-5), mono-cyte chemoattractant protein-1 (MCP-1) and stromal derived factor-1 (SDF-1) in 120 medication-naïve, physi-cally well patients with a diagnosis of GAD and comorbid personality disorder with 40 controls.48 Significantly higher levels of MCP-1 and SDF-1 were reported in both men and women, and higher CCL-5 in men but not women with a diagnosis of GAD compared with controls.48

DIsCussIOnTo our knowledge, this is the first systematic review and meta-analysis focusing on inflammatory cytokines in GAD. Using a range of databases we identified 14 studies, comprising 1188 participants with GAD and which measured 16 cytokines. We found significantly raised levels of CRP, IFN-γ and TNF-α in people with GAD compared with controls, which were findings replicated in two or more studies. A further 10 proinflammatory cytokines were reported to be significantly raised in GAD in at least one study; however, 6 of 14 studies found no difference in at least one cytokine.

Despite substantial efforts to acquire data by contacting the authors, it was only possible to conduct a meta-analysis of CRP. This identified significantly higher levels in GAD compared with controls with a small effect size (SMD=0.38), although there was evidence of significant heterogeneity across studies (I2=75%). This effect size in CRP is greater than has been reported in other anxiety disorders (PTSD: SMD=−0.14)24 or MDD (SMD=0.14)6) and is similar to that reported in schizo-phrenia (SMD=0.45).49 However, the effect size of our meta-analysis was driven by findings in poorer quality studies with small sample sizes. The two higher quality, larger studies reported a smaller effect size and no signif-icant difference between groups, respectively. As a result, further high-quality studies are required to confirm our findings of raised CRP in GAD.

Although we were only able to meta-analyse CRP, meta-analyses of different cytokines in other anxiety disorders have been conducted with larger effect sizes. A meta-analysis of inflammatory markers in PTSD iden-tified 20 studies which reported increased IL-6, IL-1β, TNF-α and IFN-γ levels, with effect sizes ranging from small (IFN-γ: SMD 0.49) to large 1.42 (IL-1β: SMD 1.42).24 However, a systematic review and meta-analysis of proinflammatory cytokines in OCD identified 12 studies,

and concluded that there was a significant reduction in IL-1β with moderate effect size (SMD=−0.60, p<0.001), and only IL-6 levels were significantly increased after subgroup analysis in medication-free adults with OCD.7 It is unclear whether this profile of inflammatory marker changes would follow a similar pattern in GAD if future studies enabled further meta-analysis.

In light of the high heterogeneity among studies, low participant numbers, and inconsistent reporting and adjustment for known confounding factors such as BMI, smoking, medication use and comorbidities, our findings should be interpreted with caution. It was not possible to analyse the cause of the degree of heterogeneity due to the paucity of studies. Other known mediators of inflam-mation24 such as physical activity, raised blood pressure and genetics were not accounted for. Furthermore, reporting of GAD severity and duration of symptoms was generally poor, preventing detailed analysis of whether inflammatory markers predicted outcomes and quality of life. We also found limitations in inclusion of specific demographics of participants with GAD. For example, despite GAD in older adults being prevalent and often treatment-resistant,50–52 only two studies included partic-ipants over the age of 65, both of which only included patients with comorbid ischaemic heart disease.

We are beginning to understand the interplay between cytokines, the immune system and mental health.1 53 At a molecular level we are aware that proinflammatory cytokines, including IFN, IL-1β and TNF, can reduce the availability of monoamines by inducing expression of presynaptic reuptake pumps and inhibiting enzymes involved in monoamine synthesis,54 linking the mono-amine theory of anxiety with inflammatory mechanisms. There is also a growing understanding of the relationship between systemic inflammation and the central nervous system (CNS).1 55 Microglial activation has been shown to be mediated by peripheral cytokines, and increased acti-vation has been found in postmortem studies of patients with MDD and schizophrenia.1 No study we identified correlated inflammatory marker changes with in vivo microglial activation imaging in GAD, and to our knowl-edge no research on postmortem microglial changes in GAD has been conducted. Increased neuronal activity has also been shown to induce inflammatory and vascular changes in the brain, suggesting that psychological stress can not only be induced by inflammation but perpetuate chronic low-grade inflammation seen in other vascular and neurodegenerative disorders.55 Understanding interactions between the CNS and immune system and identifying biomarkers of GAD offer potential for novel therapeutic approaches. The revolution of development of monoclonal-antibody therapies for inflammatory disor-ders56 raises the possibility of repurposing these medi-cations for trials in treatment-resistant GAD if specific and consistent profiles of inflammatory biomarkers are identified.

However, it remains unclear as to whether inflamma-tion plays a causal role in GAD.43 54 For example, although

on Novem

ber 6, 2020 by guest. Protected by copyright.

http://bmjopen.bm

j.com/

BM

J Open: first published as 10.1136/bm

jopen-2018-027925 on 19 July 2019. Dow

nloaded from

15Costello H, et al. BMJ Open 2019;9:e027925. doi:10.1136/bmjopen-2018-027925

Open access

IL-6 is a successful target for treatment in a number of autoimmune conditions and raised IL-6 is implicated in several psychiatric disorders, it also acts to reduce other proinflammatory cytokines such as TNF via negative feed-back and is induced by physical exercise, hyperthermia, fasting, sleep deprivation and sunlight exposure without activation of other proinflammatory cytokines.43 This raises the question as to whether inflammation in GAD is a consequence rather than a cause of symptoms. This will only be answered by large prospective longitudinal studies, better characterising the relationship between inflammation and GAD. However, remarkably our review identified only one longitudinal study of inflammation in patients with GAD that examined a cohort of adolescents until the age of 21 and only investigated CRP.

Recent studies using Mendelian randomisation in depression have suggested that cytokines such as IL-6 are causal risk factors for depression,57 and trials of immuno-therapy in psychosis are already under way.58 Our study suggests that GAD is an important candidate for future similar studies exploring causality of inflammation and potentially novel drug trials.

COnClusIOnThere is some preliminary evidence to suggest a raised inflammatory response in GAD, although it is unclear whether inflammatory cytokines play a role in the aeti-ology. GAD remains a poorly studied area of psychi-atric neuroinflammatory research compared with other mental illnesses such as MDD and schizophrenia. While we are a long way from using inflammatory cytokines as a biomarker or treatment target in GAD, current find-ings reflect inflammatory changes seen in other mental illnesses and highlight the importance of ongoing investi-gation of the role inflammation plays in the development and course of GAD. Further, methodologically consis-tent, prospective, longitudinal studies examining the mechanisms and relationship between inflammation and GAD, while accounting for known mediators of cytokine production, are required.

Contributors RH, RLG and HC were involved in the initial design of the research. EA and HC performed the data extraction of the included studies and quality analysis. HC wrote the initial draft of the manuscript and did the statistical analysis, with supervision from RLG and RH. RH, RLG and HC participated in the critical revision of the article, and all authors approved the final article.

Funding HC and EA are NIHR Academic Clinical Fellows (ACF). This research was supported by the NIHR Biomedical Research Centre at University College London/University College Hospital London.

Competing interests None declared.

Patient consent for publication Not required.

Provenance and peer review Not commissioned; externally peer reviewed.

Data sharing statement All data extracted for this systematic review and meta-analysis are available via direct contact with HC.

Open access This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given,

and indication of whether changes were made. See: https:// creativecommons. org/ licenses/ by/ 4. 0/.

rEFErEnCEs 1. Réus GZ, Fries GR, Stertz L, et al. The role of inflammation and

microglial activation in the pathophysiology of psychiatric disorders. Neuroscience 2015;300:141–54.

2. Brown AS, Vinogradov S, Kremen WS, et al. Prenatal exposure to maternal infection and executive dysfunction in adult schizophrenia. Am J Psychiatry 2009;166:683–90.

3. Benros ME, Waltoft BL, Nordentoft M, et al. Autoimmune diseases and severe infections as risk factors for mood disorders: a nationwide study. JAMA Psychiatry 2013;70:812.

4. O’dushlaine C, Rossin L, Lee PH, et al. Psychiatric genome-wide association study analyses implicate neuronal, immune and histone pathways. Nat Neurosci 2015;18:199–209.

5. Khandaker GM, Cousins L, Deakin J, et al. Inflammation and immunity in schizophrenia: implications for pathophysiology and treatment. Lancet Psychiatry 2015;2:258–70.

6. Howren MB, Lamkin DM, Suls J. Associations of depression with C-reactive protein, IL-1, and IL-6: a meta-analysis. Psychosom Med 2009;71:171–86.

7. Gray SM, Bloch MH. Systematic review of proinflammatory cytokines in obsessive-compulsive disorder. Curr Psychiatry Rep 2012;14:220–8.

8. Munkholm K, Vinberg M, Vedel Kessing L. Cytokines in bipolar disorder: a systematic review and meta-analysis. J Affect Disord 2013;144:16–27.

9. Meyer U, Feldon J. Neural basis of psychosis-related behaviour in the infection model of schizophrenia. Behav Brain Res 2009;204:322–34.

10. Goldsmith DR, Rapaport MH, Miller BJ. A meta-analysis of blood cytokine network alterations in psychiatric patients: comparisons between schizophrenia, bipolar disorder and depression. Mol Psychiatry 2016;21:1696–709.

11. Rao JS, Harry GJ, Rapoport SI, et al. Increased excitotoxicity and neuroinflammatory markers in postmortem frontal cortex from bipolar disorder patients. Mol Psychiatry 2010;15:384–92.

12. Khairova RA, Machado-Vieira R, Du J, et al. A potential role for pro-inflammatory cytokines in regulating synaptic plasticity in major depressive disorder. Int J Neuropsychopharmacol 2009;12:561.

13. Kubera M, Obuchowicz E, Goehler L, et al. In animal models, psychosocial stress-induced (neuro)inflammation, apoptosis and reduced neurogenesis are associated to the onset of depression. Prog Neuropsychopharmacol Biol Psychiatry 2011;35:744–59.

14. Furtado M, Katzman MA. Neuroinflammatory pathways in anxiety, posttraumatic stress, and obsessive compulsive disorders. Psychiatry Res 2015;229:37–48.

15. Andlin-Sobocki P, Wittchen HU. Cost of anxiety disorders in Europe. Eur J Neurol 2005;12 Suppl 1:39–44.

16. Baxter AJ, Scott KM, Vos T, et al. Global prevalence of anxiety disorders: a systematic review and meta-regression. Psychol Med 2013;43:897–910.

17. Bystritsky A. Treatment-resistant anxiety disorders. Mol Psychiatry 2006;11:805–14.

18. Bruce SE, Yonkers KA, Otto MW, et al. Influence of psychiatric comorbidity on recovery and recurrence in generalized anxiety disorder, social phobia, and panic disorder: a 12-year prospective study. Am J Psychiatry 2005;162:1179–87.

19. Baldwin D, Woods R, Lawson R, et al. Efficacy of drug treatments for generalised anxiety disorder: systematic review and meta-analysis. BMJ 2011;342:d1199.

20. Cuijpers P, Sijbrandij M, Koole S, et al. Psychological treatment of generalized anxiety disorder: a meta-analysis. Clin Psychol Rev 2014;34:130–40.

21. Härter MC, Conway KP, Merikangas KR. Associations between anxiety disorders and physical illness. Eur Arch Psychiatry Clin Neurosci 2003;253:313–20.

22. Roy-Byrne PP, Davidson KW, Kessler RC, et al. Anxiety disorders and comorbid medical illness. Gen Hosp Psychiatry 2008;30:208–25.

23. Uguz F, Akman C, Kucuksarac S, et al. Anti-tumor necrosis factor-alpha therapy is associated with less frequent mood and anxiety disorders in patients with rheumatoid arthritis. Psychiatry Clin Neurosci 2009;63:50–5.

24. Passos IC, Vasconcelos-Moreno MP, Costa LG, et al. Inflammatory markers in post-traumatic stress disorder: a systematic review, meta-analysis, and meta-regression. Lancet Psychiatry 2015;2:1002–12.

on Novem

ber 6, 2020 by guest. Protected by copyright.

http://bmjopen.bm

j.com/

BM

J Open: first published as 10.1136/bm

jopen-2018-027925 on 19 July 2019. Dow

nloaded from

16 Costello H, et al. BMJ Open 2019;9:e027925. doi:10.1136/bmjopen-2018-027925

Open access

25. Michopoulos V, Powers A, Gillespie CF, et al. Inflammation in Fear- and Anxiety-Based Disorders: PTSD, GAD, and Beyond. Neuropsychopharmacology 2017;42:254–70.

26. Liberati A, Altman DG, Tetzlaff J, et al. The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate health care interventions: explanation and elaboration. J Clin Epidemiol 2009;62:e1–34.

27. Spitzer R, Williams J, Gibbon M, et al. Structured Clinical Interview for DSM-IV. Encyclopedia of Behavioral Medicine 1994.

28. Wells G, Shea B, O’Connell D, et al. The Newcastle-Ottawa Scale (NOS) for Assessing the Quality of Nonrandomized Studies in Meta-Analyses. 2013.

29. Higgins JP, Thompson SG, Deeks JJ, et al. Measuring inconsistency in meta-analyses. BMJ 2003;327:557–60.

30. Kitahara CM, Trabert B, Katki HA, et al. Body mass index, physical activity, and serum markers of inflammation, immunity, and insulin resistance. Cancer Epidemiol Biomarkers Prev 2014;23:2840–9.

31. Petrovsky N, McNair P, Harrison LC. Diurnal rhythms of pro-inflammatory cytokines: regulation by plasma cortisol and therapeutic implications. Cytokine 1998;10:307–12.

32. Akdis M, Aab A, Altunbulakli C, et al. Interleukins (from IL-1 to IL-38), interferons, transforming growth factor β, and TNF-α: Receptors, functions, and roles in diseases. J Allergy Clin Immunol 2016.

33. Khandaker GM, Zammit S, Lewis G, et al. Association between serum C-reactive protein and DSM-IV generalized anxiety disorder in adolescence: Findings from the ALSPAC cohort. Neurobiol Stress 2016;4:55–61.

34. Nayek S, Ghosh S. A comparative study of serum C-Reactive protein in patients with generalised anxiety disorder and depression. Indian J Psychiatry 2017;59:S217.

35. Tang Z, Ye G, Chen X, et al. Peripheral proinflammatory cytokines in Chinese patients with generalised anxiety disorder. J Affect Disord 2018;225:593–8.

36. Vogelzangs N, Beekman AT, de Jonge P, et al. Anxiety disorders and inflammation in a large adult cohort. Transl Psychiatry 2013;3:e249.

37. Sterne JA, Sutton AJ, Ioannidis JP, et al. Recommendations for examining and interpreting funnel plot asymmetry in meta-analyses of randomised controlled trials. BMJ 2011;343:d4002.

38. Bankier B, Barajas J, Martinez-Rumayor A, et al. Association between C-reactive protein and generalized anxiety disorder in stable coronary heart disease patients. Eur Heart J 2008;29:2212–7.

39. Copeland WE, Shanahan L, Worthman C, et al. Generalized anxiety and C-reactive protein levels: a prospective, longitudinal analysis. Psychol Med 2012;42:2641–50.

40. Korkeila J, Runsten S, Ollikainen S, et al. Generalized anxiety disorder and immunity markers in a stratified population sample. Eur Psychiatry 2010:25.

41. Zahm JL. Generalized anxiety disorder and inflammatory biomarkers in coronary heart disease: Sex-specific effects. Diss Abstr Int Sect B Sci Eng 2018;78.

42. De Berardis D, Serroni N, Campanella D, et al. Alexithymia, suicide ideation, c-reactive protein, and serum lipid levels among outpatients with generalized anxiety disorder. Arch Suicide Res 2017;21:100–12.

43. Raison CL, Knight JM, Pariante C. Interleukin (IL)-6: A good kid hanging out with bad friends (and why sauna is good for health). Brain Behav Immun 2018;73:1–2.

44. Hoge EA, Bui E, Palitz SA, et al. The effect of mindfulness meditation training on biological acute stress responses in generalized anxiety disorder. Psychiatry Res 2018;262:328–32.

45. Yang CJ, Liu D, Xu ZS, et al. The pro-inflammatory cytokines, salivary cortisol and alpha-amylase are associated with generalized anxiety disorder (GAD) in patients with asthma. Neurosci Lett 2017;656:15–21.

46. Tofani T, Di Cesare Mannelli L, Zanardelli M, et al. P.1.f.003 An immunologic profile study in drug-näive generalized anxiety non depressed patients: a pilot study. European Neuropsychopharmacology 2015;25:S226.

47. Hou R, Garner M, Holmes C, et al. Peripheral inflammatory cytokines and immune balance in Generalised Anxiety Disorder: Case-controlled study. Brain Behav Immun 2017;62:212–8.

48. Ogłodek EA, Szota AM, Just MJ, et al. The MCP-1, CCL-5 and SDF-1 chemokines as pro-inflammatory markers in generalized anxiety disorder and personality disorders. Pharmacol Rep 2015;67:85–9.

49. Miller BJ, Culpepper N, Rapaport MH. C-reactive protein levels in schizophrenia: a review and meta-analysis. Clin Schizophr Relat Psychoses 2014;7:223-30.

50. Byers AL, Yaffe K, Covinsky KE, et al. High occurrence of mood and anxiety disorders among older adults: The National Comorbidity Survey Replication. Arch Gen Psychiatry 2010;67:489-96.

51. Wolitzky-Taylor KB, Castriotta N, Lenze EJ, et al. Anxiety disorders in older adults: a comprehensive review. Depress Anxiety 2010;27:190–211.

52. Wetherell JL, Afari N, Ayers CR, et al. Acceptance and commitment therapy for generalized anxiety disorder in older adults: a preliminary report. Behav Ther 2011;42:127–34.

53. Felger JC, Lotrich FE. Inflammatory cytokines in depression: neurobiological mechanisms and therapeutic implications. Neuroscience 2013;246:199–229.

54. Miller AH, Raison CL. The role of inflammation in depression: from evolutionary imperative to modern treatment target. Nat Rev Immunol 2016;16:22–34.

55. Xanthos DN, Sandkühler J. Neurogenic neuroinflammation: inflammatory CNS reactions in response to neuronal activity. Nat Rev Neurosci 2014;15:43–53.

56. Reichert JM, Rosensweig CJ, Faden LB, et al. Monoclonal antibody successes in the clinic. Nat Biotechnol 2005;23:1073–8.

57. Khandaker GM, Zuber V, Rees JMB, et al. Shared mechanisms between coronary heart disease and depression: findings from a large UK general population-based cohort. Mol Psychiatry 2019.

58. Miller BJ, Buckley PF. The Case for Adjunctive Monoclonal Antibody Immunotherapy in Schizophrenia. Psychiatr Clin North Am 2016;39:187–98.

on Novem

ber 6, 2020 by guest. Protected by copyright.

http://bmjopen.bm

j.com/

BM

J Open: first published as 10.1136/bm

jopen-2018-027925 on 19 July 2019. Dow

nloaded from